Survival models for bone marrow transplantation data.

Bone Marrow Transplants (BMT) are a standard treatment for acute Leukemia. Prediction for recovery could depend on previously known risk factors, such as patient's disease group and time to transplantation. In this paper, different procedures were applied and tested in order to fit the data of BMT disease free survival time. Fitting this data to multiple possible models such as non-parametric and semiparametric survival models in order to find the most adequate one based on several variables in the data sets. For non-parametric methods, a Kaplan-Meier (KM) estimator is used in order to estimate survival functions based on the disease group (g) and four hospital (Z9). Form the semiparametric methods a Cox model was used with different covariates. Followed by some hypothesis testing to analyses the goodness of fit. It was found that Cox model is suitable for this data set with special covariates that is disease group (g), hospital (Z9) and MTX prophylactic (Z10) with respect to LRT, Wald, Score tests and the Cox-Snell residual. Moreover, some of the assumptions of Cox model were satisfied. [ABSTRACT FROM AUTHOR]